Logos52
wiki / Syntheses / Minimally Viable Learning System

Minimally Viable Learning System

synthesis updated 2026-08-14

Minimally Viable Learning System

A minimum learning system is a loop small enough to run on a tired day and still do two jobs. The jobs are to build a structure so knowledge has somewhere to sit, then reconstruct it later so it becomes usable instead of decorative. The system becomes too heavy when the session is spent managing the method instead of learning through it.

The house names for those jobs come after the jobs. Bear Hunter System encodes usable structure: a rough frame, the questions the material must answer, a working map of connections, then a cut-down structure that survives without the source. Spaced Interleaved Retrieval retrieves that structure, spaces the returns, mixes topics and forms, and repairs what will not rebuild.

The five questions

A session has a spine when five questions have answers.

  1. Was there enough preparation to know what to look for?
  2. Was a usable structure built?
  3. Can that structure be reconstructed without looking?
  4. Did retrieval reveal a gap?
  5. Was the gap repaired?
StepQuestionOutput
PrestudyEnough preparation to know what to look for?A rough frame and a few questions before the main pass
Bear Hunter SystemWas a usable structure built?A map that can be pruned until it survives without the source
Spaced Interleaved RetrievalCan it be reconstructed without looking? Did a gap show?A list of what would not rebuild
RepairWas the gap repaired?One weak point closed in the same session
RepeatStill the two jobs?The loop again, not a new method

That is the spine. Frame, encode, retrieve, detect gaps, repair. Everything else supports, diagnoses, or extends it.

The same sequence as a study block: Prestudy, then Bear Hunter System, then Spaced Interleaved Retrieval, then repair one weak point, then ask which dimension limited the result. The word for later additions is upgrades. Kolbs Experiential Cycle, Marginal Gains, Focus Management: How to Enter & Recover Inside a Work Block, and The Technique Is Only as Good as the Thinking It Produces go on when they improve the core loop. They go off when they make the system heavier.

What makes it too heavy

Six failures, each with a repair.

FailureWhat it looks likeRepair
OvercompressionA vague loop that has lost both jobsPut the five questions back. If Bear Hunter System and Spaced Interleaved Retrieval have become synonyms for “study,” the spine is gone
Technique sprawlAnother method added because the last one felt incompleteRun the five questions on the hour that already happened. Add only what answers one of them
Dashboard driftTracking the system instead of using itIf the session produced a nicer tracker and no reconstructed structure, drop the tracker
Forced integrationA study model pushed onto a build, or a build model pushed onto a studyKeep the two jobs for study. Do not require a third engine
Archive illusionPages stored, nothing retrievedA page counts when the next question, retrieval, or action changes
Dimension abstractionImproving all five lenses as if they were the coreRun the two jobs, then ask which lens limited the result

The five dimensions improve the minimum by diagnosing where the two jobs went weak. They are not a second core. Deep Processing asks whether the encoding pass built relationships or only organised notes. Retrieval asks whether the return forced reconstruction or only recognition. Self-Regulation asks whether failure was visible mid-session. Self-Management asks whether the conditions existed. Mindset asks how difficulty was read. Run the two jobs, then ask which of those limited the result. Do not improve all five separately.

Dimensions of Learning is the home of those five lenses. Prestudy, BHS, and SIR: Turning Information into Usable Structure owns the loop at full depth. This page owns the weight limit.

The case against the page is that a thin loop will feel unserious, and a tired-day test will be read as permission to skip the encoding pass. The price of the limit is leaving methods on the table. Quit if the session is spent on the system. The checkable expectation is five answered questions and one repaired gap.

After the weight is visible

The two jobs are still the test. A loop that still works on a tired day is complete enough when the five questions still have answers. That is enough to stop consuming information that never becomes usable. If the hour went into the system, drop back to the five questions.

Open Questions

  • Can Bear Hunter System and Spaced Interleaved Retrieval stay the stable minimum, while Agentic Engineering and the knowledge base plug in only when they improve usable structure?
  • Agentic Engineering is adjacent. A possible fit is encode the problem, retrieve the constraints, let the build workflow implement, then run Kolbs if a repeatable bottleneck appeared. The danger is forcing a study model onto a build model. It may use the two jobs as support and still need its own workflow.
  • The knowledge base has three roles: archive, thinking partner, retrieval surface. A page earns its keep when it creates reusable structure, is retrieved during a real question, and changes the next move. Does it improve thought, or only store it?

Sources

  • Craik, F. I. M., & Lockhart, R. S. (1972). Levels of processing: A framework for memory research. Journal of Verbal Learning and Verbal Behavior, 11(6), 671–684. Depth of the encoding operation is what the first job is.
  • Roediger, H. L., & Karpicke, J. D. (2006). Test-enhanced learning. Psychological Science, 17(3), 249–255. Reconstruction is the second job.
  • Cepeda, N. J., Pashler, H., Vul, E., Wixted, J. T., & Rohrer, D. (2006). Distributed practice in verbal recall tasks: A review and quantitative synthesis. Psychological Bulletin, 132(3), 354–380. Spacing is how the second job stays alive.